Molecular dynamics simulations of Piezo1 channel opening by increases in membrane tension
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Article
Molecular dynamics simulations of Piezo1 channel
opening by increases in membrane tension
Dario De Vecchis,1 David J. Beech,1 and Antreas C. Kalli1,2,*
1
Leeds Institute of Cardiovascular and Metabolic Medicine, School of Medicine and 2Astbury Centre for Structural Molecular Biology, University
of Leeds, Leeds, United Kingdom
ABSTRACT Piezo1 is a mechanosensitive channel involved in many cellular functions and responsible for sensing shear
stress and pressure forces in cells. Piezo1 has a unique trilobed topology with a curved membrane region in the closed state.
It has been suggested that upon activation Piezo1 adopts a flattened conformation, but the molecular and structural changes
underpinning the Piezo1 gating and opening mechanisms and how the channel senses forces in the membrane remain elusive.
Here, we used molecular dynamics simulations to reveal the structural rearrangements that occur when Piezo1 moves from a
closed to an open state in response to increased mechanical tension applied to a model membrane. We find that membrane
stretching causes Piezo1 to flatten and expand its blade region, resulting in tilting and lateral movement of the pore-lining trans-
membrane helices 37 and 38. This is associated with the opening of the channel and movement of lipids out of the pore region.
Our results reveal that because of the rather loose packing of Piezo1 pore region, movement of the lipids outside the pore region
is critical for the opening of the pore. Our simulations also suggest synchronous flattening of the Piezo1 blades during Piezo1
activation. The flattened structure lifts the C-terminal extracellular domain up, exposing it more to the extracellular space. Our
studies support the idea that it is the blade region of Piezo1 that senses tension in the membrane because pore opening failed in
the absence of the blades. Additionally, our simulations reveal that upon opening, water molecules occupy lateral fenestrations
in the cytosolic region of Piezo1, which might be likely paths for ion permeation. Our results provide a model for how mechanical
force opens the Piezo1 channel and thus how it might couple mechanical force to biological response.
SIGNIFICANCE Cells have sophisticated molecular machines that help them to respond to external mechanical forces.
One of these machines is the membrane protein Piezo1. Piezo1 has a critical role in the circulatory system and tissue
development and is associated with human diseases and disorders such as hemolytic anemia. For this reason, it is critical
to understand how Piezo1 functions at the molecular level. In this study, we use computer simulations to study Piezo1
activation. Our studies suggest that Piezo1 adapts to the stretch of the membrane by flattening its structure and reducing
its membrane indentation. Hence, we propose a model that explains the steps by which Piezo1 channels may adapt and
open in response to increased membrane tension.
INTRODUCTION Structural data show that Piezo1 is a trimer (9–12) with a
trilobed topology and a curved membrane region that forms
Piezo1 is a critical mechanical sensor that is found in endo-
the blades (9–12). On the intracellular side, each Piezo1
thelial cells, red blood cells, and other cell types (1–3). It
subunit has a beam region (i.e., an extended a-helix) that
plays a critical role in the circulatory system and tissue
connects the periphery of Piezo1 with its central region.
development. Mutations in Piezo1 are linked to human dis-
The three Piezo1 subunits converge to a central pore region
eases such as lymphedema (2,4) and hematological disor-
with a C-terminal extracellular domain (CED) between each
ders such as hemolytic anemia (5) and resistance to
of the last two transmembrane (TM) helices. The pore re-
malaria (6). Its main function is to permeate ions, including
gion of Piezo1 is surrounded by three blades. Each blade
Naþ and Ca2þ in response to mechanical stimuli (3,7,8).
consists of nine four-a-helix bundles; the three N-terminal
bundles are not present in this structural data for Piezo1.
Submitted August 3, 2020, and accepted for publication February 1, 2021. It has been shown that because of its unique curved mem-
*Correspondence: a.kalli@leeds.ac.uk brane region (i.e., the blade region), Piezo1 induces a dome
Editor: Marta Filizola. in the membrane (11) when Piezo1 is in its closed state.
https://doi.org/10.1016/j.bpj.2021.02.006
Ó 2021 Biophysical Society.
This is an open access article under the CC BY license (http://
creativecommons.org/licenses/by/4.0/).
1510 Biophysical Journal 120, 1510–1521, April 20, 2021
Piezo1 opening by membrane tension
Mechanical calculations suggested that in addition to the were added with MODELER (v 9.19) (25), and the loop refinement tool
local dome, Piezo1 creates a wider footprint in the mem- (26) was used to remove a knot in one chain between residues 2066 and
2074. The best loop was selected out of 10 candidates according to the
brane that extends beyond the Piezo1 radius (13). This discrete optimized protein energy method (27). The final Piezo1 model
extended footprint may regulate Piezo1 sensitivity to me- does not include the first 576 residues because they are not present in the
chanical force (13). Upon activation, Piezo1 was shown to template and residues 718–781, 1366–1492, 1579–1654, and 1808–1951,
adopt a flatter conformation (14,15); however, the molecular which are exposed to the cytosol. Therefore, each chain is composed by
details of Piezo1’s active conformation as well as its inter- five nonoverlapping fragments: residues 577–717, 782–1365, 1493–1578,
1655–1807, and 1952–2547. This Piezo1 model was also used to prepare
mediate structural states during activation are unknown. the two blade-free constructs comprising of residues 1973–2547
Flattening of Piezo1 during activation may be the result of (Piezo1D1973) and 2104–2547 (Piezo1D2104). Piezo1D1973 has only the last
the in-plane area expansion (11), as also seen for MscL four-helical bundle which is the closest bundle to the channel pore. Pie-
mechanosensitive channel (16). Hypotheses for Piezo1 zo1D2104 is a complete blade-free Piezo1.
gating include the ‘‘force-from-lipids’’ principle (17,18),
which suggests that Piezo1 senses mechanical forces
through the bilayer (1,19) and a direct involvement of the Coarse-grained simulations
cytoskeleton as well as the extracellular matrix in Piezo1 The sizes and compositions of all the simulated systems are listed in Table
activation (20). However, the molecular and structural S1. For all the Piezo1 models described above, the protein coordinates were
changes underpinning the Piezo1 gating mechanism and converted to a coarse-grained (CG) resolution and energy minimized in a
vacuum with GROMACS 2016 (Fig. S1) (28). Fig. S1 shows that despite
how the channel senses forces in the membrane remain the minimization in a vacuum, no conformational changes were observed
unknown. within Piezo1 during the minimization step. The CG-MD simulation was
In silico methodologies have been successfully used to essential to equilibrate the lipid bilayer around the protein and reconstitute
study membrane mechanics under stress (21,22), including the Piezo1 membrane indentation. The CG-MD simulations were per-
for Piezo1 (15). However, although ms-long atomistic simu- formed using the Martini 2.2 force field (29) and GROMACS (28). To
model the protein secondary and tertiary structure, an elastic network model
lation of a partial Piezo1 structure at a physiological nega- with a cutoff distance of 7 Å was used. For the equilibration simulation
tive pressure detected local fenestrations, such simulation Piezo1 model was inserted in a complex asymmetric bilayer using the
did not record a full opening event (15). Indeed, the activa- INSert membrANE tool (30). Three independent system assembly steps
tion time of mechanosensitive channels is in the order of and as many CG equilibrations were carried out, the first of which were
milliseconds (23,24), which is currently challenging with used in this study (see Table S1, Equil 1). During the CG-MD simulations,
the protein was position restrained, and thus, no changes within the protein
atomistic molecular dynamics (MD) simulations. were allowed. With the exception of the simulation carried out in a 1,2-di-
In this study, we used MD simulations and comparison lauroyl-sn-glycero-3-phosphocholine (DLPC) (Piezo1DLPC) bilayer, the
with available knowledge to suggest the structural rear- composition of the model bilayer was the following: for the outer leaflet,
rangements that occur when Piezo1 progresses from a 1-palmitoyl-2-oleyl-phosphtidylcholine (POPC) 55%, sphingomyelin
closed to an open state when increasing mechanical tensions (SM) 5%, 1-palmitoyl-2-oleyl-phosphtidylethanolamine (POPE) 20%,
and cholesterol 20%. For the inner leaflet, POPC 50%, POPE 20%, 1-pal-
are applied to a membrane bilayer. Our results suggest a mitoyl-2-oleyl-phosphtidylserine (POPS) 5%, cholesterol 20%, and phos-
relationship between membrane mechanics and Piezo1 phatidylinositol 4,5-bisphosphate (PIP2) 5%. The system was neutralized
structural rearrangements and show that upon stretching with a 150 mM concentration of NaCl. The model was further energy mini-
Piezo1 moves to a flatter stable conformation propelled by mized and subsequently equilibrated for 500 ns with the protein particles
the simultaneous expansion of its blade regions. We also restrained (1000 kJ $ mol1 $ nm2) to allow the membrane bilayer to
equilibrate around the model. The time step for all the CG simulations
show the critical role that the blade regions play in sensing was 20 fs. The equilibration was performed at 323 K, with protein, lipids,
tension forces in the membrane. Piezo1 flattening and and solvent separately coupled to an external bath using the v-rescale ther-
expansion result in an increase of the in-plane area and mostat (31) (coupling constant of 1.0). The temperature of 323 K is above
changes in the beam helix tilt angle. Our Piezo1 flat open the transition temperatures of all lipid species in the systems, therefore
model also shows solvated lateral fenestrations in the cyto- avoiding the lipids to undergo phase transitions to the gel phase. Pressure
was maintained at 1 bar (coupling constant of 1.0) with semi-isotropic con-
solic region of Piezo1, which might be likely paths for ion ditions and compressibility of 3 106 using the Berendsen barostat (32).
permeation. These events lead in turn to the opening of Lennard-Jones and Coulombic interactions were shifted to 0 between 9 and
the channel and to the migration of lipids outside of the 12 Å and between 0 and 12 Å, respectively.
initially occupied Piezo1 central pore region, suggesting
structural changes that happen during Piezo1 mechanical
activation. Atomistic simulations
After the CG-MD equilibration (Equil 1 in Table S1), two POPC molecules
were removed because the headgroup was trapped between Piezo1 TM bun-
dles. Moreover, to restore membrane asymmetry, three molecules of PIP2
MATERIALS AND METHODS
were removed because they flipped to the outer leaflet. The number of lipids
Modeling the Piezo1 trimer removed is extremely small compared with the number of lipids in each
leaflet (1300 phospholipids and 300 cholesterol molecules in each
Structural data were obtained from the cryogenic electron microscopy leaflet), and thus, it did not introduce any bias during our simulations.
(cryo-EM) structure Protein Data Bank, PDB: 6B3R (11). Missing residues The system obtained was minimized and further converted to atomistic
Biophysical Journal 120, 1510–1521, April 20, 2021 1511
De Vecchis et al.
resolution as described in (33). A moderate concentration of 3.0 mM of lated as follows: the program visual molecular dynamics 1.9.3 (VMD) (40)
Ca2þ (i.e., 50 ions) was added to the box, and neutrality was restored (http://www.ks.uiuc.edu/Research/vmd/) was used to calculate distances
with counterions. The Ca2þ was added because Piezo1 was also shown to between the Ca atom from residue Ala641, located within the first bundle
be permeable to Ca2þ. The obtained system was energy minimized and sub- embedded in the bilayer from each Piezo1 chain and for each trajectory.
sequently equilibrated in four NPT ensemble runs of 20,000 steps each with The distance was then used to calculate the radius r of the circumscribed
an increasing time step from 0.2 to 2 fs, and the protein particles were circle and considering the Piezo1 triskelion as the vertices of a scalene tri-
restrained (1000 kJ $ mol1 $ nm2). A further equilibration step of 5 ns angle that is planar to the surface of the membrane bilayer using an in-house
with a time step of 1 fs and Ca atoms restrained (1000 kJ $ mol1 $ script according to Eq. 1.
nm2) was performed to relax the Piezo1 model embedded in a highly
curved bilayer (i.e., the membrane indentation). Very quickly in the equil- ða b cÞ
r ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi: (1)
ibration, quantities such as box size and volume reached a value and re- ða þ b þ cÞðb þ c aÞðc þ a bÞða þ b cÞ
mained in that value for the rest of the equilibration. Stretch-induced
conformational changes in the Piezo1 model were investigated by NPT
ensemble unrestrained simulations of 50 ns in which the bilayer plane The values a, b, and c are the sides of the scalene triangle. The area of the
(x-y plane) pressure was varied semi-isotropically between 40, 30, circumference was then calculated as p r2. For the radius r and the Aproj,
20, 5, and þ1 bar, whereas the pressure in bilayer normal (z) direction the standard deviation was calculated. The average partial density was
was kept at þ1 bar. The þ1- and 30-bar systems were further extended to calculated using the GROMACS 2016 (28) tool gmx density. The two-
100 ns. A similar approach was used to investigate structural transitions dimensional density maps for lipids and cholesterol were calculated using
for the TREK-2 mechanosensitive channel (34). All the atomistic systems gmx densmap. The tilting angle relative to the bilayer normal and distance
were simulated using GROMACS 2016 (28) with CHARMM36 force field from the Piezo1 pore of the Piezo1 beams, pore-lining helices 38 and 37,
(35) and a 2-fs time step. A Berendsen semi-isotropic pressure coupling were calculated using gmx bundle and gmx distance tools from GROMACS
(32) at 1 bar was used during all the equilibration phases. The Parrinello- 2016 (28). For all calculations, the Piezo1 pore region comprised the same
Rahman barostat (36) was further used for the stretch-induced simulations. residues that were used for the calculation of the cavity (see below). The
All simulations were performed at 323 K, with protein, lipids, and solvent tool trj_cavity (41) (https://sourceforge.net/projects/trjcavity/) was used
coupled to an external bath using the v-rescale thermostat. Such tempera- for the calculation of the volume of the Piezo1 pore cavity as well as for
ture was also used in simulations of other mechanosensitive channels the water molecules that overlap to it. The considered residues were
(34). Long-range electrostatics were managed using the particle-mesh 2450–2547, 2214–2224, and 2326–2334. The trajectories were fitted on
Ewald method (37). The LINCS algorithm was used to constrain bond the Ca of the considered residues before the calculation and options dim
lengths (38). 4 (i.e., the degree of how buried they were) and spacing 1.4 (i.e., the size
The same protocols described above were followed for all the all-atom of the grid voxel in Å) were used. Salt bridges were calculated with
simulations (see Table S1), with the exception of the Piezo140.rep1 and VMD 1.9.3 (40) using a 6-Å cutoff and, when indicated, without consid-
Piezo140.rep2 systems. For these systems, the protein coordinates from ering the first 30 ns of simulation. For the calculation of the area per lipid
the last frame from each of the two repeat simulations at 40 bar were (APL), the tool FATSLiM (42) (http://fatslim.github.io/) was used. The
inserted in a model asymmetric membrane by using the CHARMM- lateral pressure PL was calculated as explained for the TREK-2 mechano-
GUI webserver (39). The protein was embedded in a bilayer with the sensitive channel (34). Molecular graphics were done using VMD 1.9.3
same composition as described above. To reconstitute the Piezo1 open (40). Data were plotted using Grace (https://plasma-gate.weizmann.ac.il/
pore, lipids present in the channel mouth were removed to allow the water Grace/).
to flow and fill the Piezo1 open pore. For the Piezo140.rep1, we removed
four molecules of cholesterol, seven of POPC, and three of POPE over a
total of 956, 2516, and 957 lipids, respectively. For the Piezo140.rep2, we RESULTS AND DISCUSSION
removed five molecules of cholesterol, eight of POPC, and two of POPE
over a total of 951, 2504, and 954 lipids, respectively. The same concen- To investigate how mechanical forces act on the Piezo1
trations of Ca2þ were added to each system as above (3.0 mM), subse- channel, we simulated the recent Piezo1 structure solved
quently neutralized with counterions, and further minimized. For each by cryo-EM in a model membrane using a serial multiscale
system, two NPT equilibrations were performed of 2 and 8 ns with 1
and 2 fs, respectively. The systems consisted of 3 million atoms (see
MD simulation approach. Missing residues were added to
Table S1). Protein backbone atoms were restrained in both equilibrations the structure before the simulations (see Materials and
(1000 kJ $ mol1 $ nm2), whereas only in the first equilibration, the methods). Initial CG-MD simulations of 500 ns was carried
phosphorous atoms from the phospholipids and the oxygen atoms from out to equilibrate the bilayer around Piezo1 and observe the
the cholesterol were restrained (1000 kJ $ mol1 $ nm2) to allow the wa- unique Piezo1 footprint suggested previously (11,14,43).
ter to fill the Piezo1 open pore once the lipids were removed from the
channel mouth (see above). Piezo140.rep1 and Piezo140.rep2 systems
The asymmetric bilayer used in our simulations mimics
(see Table S1) were run at þ1 bar for 100 ns as described above for the the native endothelial membrane (44) (see Materials and
other systems. methods), which is one of the systems that Piezo1 functions
in physiologically (3,45). Three independent CG system as-
semblies followed by three independent equilibration runs
Analysis were carried out (see Materials and methods), and in all
To calculate the area of the box, the tool gmx energy from GROMACS 2016 cases, the timescale of the CG simulations (500 ns) was
(28) was used to extract the x-side and the y-side of the simulated box for enough for the dome to be formed around Piezo1 curved
each frame. The area was subsequently calculated by multiplying these blades with a depth of 6 nm in agreement with other
values as x-side y-side. The root mean-square deviation (RMSD) was
studies (11). Although the CG approach was critical for
calculated on the protein Ca using the gmx rms tool from GROMACS
2016 (28). For each system the initial coordinates were considered as a
creating the Piezo1/membrane system, the elastic network
reference structure. All the errors shown are the standard deviation. The implemented in the CG approach does not allow us to fully
in-plane projection of the Piezo1 membrane indentation (Aproj) was calcu- study the Piezo1 pressure-dependent dynamical responses.
1512 Biophysical Journal 120, 1510–1521, April 20, 2021
Piezo1 opening by membrane tension
For this reason, the system was backmapped to an atomistic it seems that Piezo1 blades sense the force that comes
resolution and further simulated in independent repeat sim- from the bilayer.
ulations applying pressure to the bilayer (i.e., along x and y Fig. 1 B shows the correlation between the area of the
axes) ranging from the native-like pressure of þ1 to 40 simulated box and the RMSD calculated using the Ca atoms
bar, as described in Materials and methods. Under these of the blades, demonstrating that they both increase until the
conditions, we were able to calculate the lateral pressure bilayer stops expanding. RMSD analysis showed that
(PL) applied on the bilayer, which is 14.2 mN/m for the although the blade bundles remain intact during stretching,
5 bar, 45.8 mN/m for the 20 bar, 59 mN/m for the they are the major contributors to the RMSD drift, whereas
30 bar, and 67.8 5 0.35 mN/m for the 40 bar system. the RMSD calculated over the Ca atoms of the CED is com-
The tension required for the half-maximal activation of parable between the different applied negative pressures
Piezo1 is 2.7 5 0.1 or 4.7 5 0.3 mN/m for outside-in (Fig. 1 C). In the bilayer without tension (þ1 bar), Piezo1
and inside-out patches (46). The tensions in our simulations maintained its curved shape even though the simulation
are higher than the experimentally derived half-maximal was twice as long (i.e., 100 ns) compared with some of
values and above the membrane rupture limit, but they the other simulations with stretch applied to the bilayer.
were used to enable shorter timescales accessible for atom- This shows that the curved shape of Piezo1 blades forces
istic simulations. the bilayer to adopt a curved trilobed shape. It also confirms
that our CG simulations were sufficient to create a stable
dome around Piezo1 when no tension was applied to the
Tension causes Piezo1 to flatten and expand
bilayer. Interestingly, although Piezo1 reaches a flat confor-
The different pressures cause expansion of the simulated mation in both the 30- and 40-bar systems, larger
box to a different degree with the membrane in the system conformational changes with more opening of the pore
adapting to each pressure. This results in two main mem- were observed in the 40-bar simulations. Even when the
brane perturbations: an increase of the APL in both leaflets 30-bar systems were extended up to 100 ns (compared
for all systems (Fig. S2 A) and a thinning of the membrane with 50 ns of the 40 bar), the structural changes in this sys-
bilayer (Fig. S2 B); the APL and the membrane thickness for tem were smaller (Fig. 1, A and B). It should also be noted
the system with the highest tension in the bilayer (i.e., 40 that although small, the 40- and 30-bar simulations have
bar) were 1 nm2 and 2.7 nm, respectively (Fig. S2). A a somewhat different effect on the membrane topology. For
similar approach revealed that an increase of the APL re- example, the thinning observed in the 40-bar simulations
sulted in the transition of the TREK-2 mechanosensitive are somewhat larger compared with 30 bar, and the APL
channel from a down to an up conformation (34). We note is somewhat smaller in the 30-bar simulation. These
that in all tensions, within the simulated time, the bilayer changes, although small, result in different changes within
is still in an intact state (i.e., no water moves through the the protein. Indeed, our simulations of Piezo1 at 30 and
bilayer). 40 bar do not have exactly the same effect on the protein.
Our analysis above shows that the APL reaches a plateau Piezo1 simulated at 30 bar does not reach a complete flat
for both leaflets within the first 10 ns of the simulation conformation as observed for the 40-bar simulations. The
(Fig. S2). Two of our systems, the 30 bar and the þ1 RMSD of the blades (Fig. 1 C) and the projected area (Fig. 1
bar, were extended up to 100 ns. In both cases, the analysis D) do not have the same values. Additionally, in the 40-bar
confirms that no further change in the APL occurs with simulations, the beam helices reached almost 90 (Fig. 2 B),
increased time (Fig. S2). and the CED becomes more exposed (Fig. 2 A).
The Piezo1 structure showed a remarkable adaptation to a As indicated above, the structure we simulated is missing
stretched bilayer, with its N-terminal blades remaining its three N-terminal four bundles because these parts were
embedded within the bilayer during all simulations and ex- not resolved in cryo-EM studies, possibility because of their
tending toward flatter conformations (Fig. 1 A). This result high flexibility. Because of the importance of the blade
is in good agreement with findings by Lin et al., who used structures in the mechanosensitivity of Piezo1, this partial
atomic force microscopy to show that application of me- Piezo1 structure (i.e., with fewer N-terminal blade bundles)
chanical force results in flatter conformations of Piezo1 might not have the same absolute force sensitivity as the
(14). The degree of flattening was different for the different full-length Piezo1 structure. This could explain why higher
tensions, with the system having the highest tension result- membrane tension values were required compared with
ing in an almost flat Piezo1 structure (Fig. 1 A). This sug- those suggested from studies of Piezo1 overexpressed in
gests that Piezo1 flattening follows the bilayer flattening HEK293t cells (46).
and not the contrary because within the timescales of our We investigated further if the large conformational
simulations, the channel seems to adapt its curvature to change we observed for the blades under large tension
accommodate the changes in the bilayer that result from are needed for Piezo1 to sense mechanical force. To this
different tensions. Therefore, we suggest that it is not the end, we simulated two Piezo1 blade-free constructs:
Piezo1 opening leading to flattening of the membrane, but the Piezo1D1973–2547, which consists of the four-helical
Biophysical Journal 120, 1510–1521, April 20, 2021 1513
De Vecchis et al. FIGURE 1 Piezo1 channel flattens, and its CED becomes exposed in response to increased membrane tension. For each system, the applied negative pres- sure in each simulation is indicated with different colors: black, 1 bar; red, 5 bar; green, 20 bar; blue, 30 bar; and orange, 40 bar. (A) Shown are snapshots of the last frame from simulations of Piezo1 embedded in a model membrane bilayer. (B) Given is the correlation between the area, i.e., x-side y-side, of the simulated box (Abox) and the RMSD calculated using the protein Ca of the Piezo1 blades. (C) Shown is the RMSD calculated using the protein Ca of the CED, dashed lines (lower values), and of the rest of the protein (i.e., the Piezo1 blades) and straight lines (higher values). (D) Shown is the Aproj (indicated by a dashed blue circle in A as a function of time for all the systems; see Materials and methods). The timeline for the Piezo1DLPC, Piezo140.rep1, and Piezo140.rep2 systems (values translated by 50 ns to 100* ns for clarity) is shown in purple, magenta, and dark green, respectively. For the two repeat simulations at 40 bar, the timelines are colored in orange (first repeat) and dark orange (second repeat). The transparent regions show the standard deviation. To see this figure in color, go online. bundle that is the most proximal to the pore, and the is the region of Piezo1 that senses mechanical stress (10). Piezo1D2104–2547, which does not have any of the nine Without this region, Piezo1 does not seem to be able to four-helical bundles (see Materials and methods). Both sense mechanical force. constructs were simulated at 40-bar pressure for 100 ns Upon flattening of its blades, Piezo1 expands, thus each to investigate if the same tension that we saw to increasing the Aproj (Fig. 1 A). During this process, no un- open Piezo1 would also open a Piezo1 blade-free channel. folding events within the protein were evident. This is in Piezo1D1973–2547 and Piezo1D2104–2547 have a similar good agreement with recent studies that also suggested an behavior and did not open at 40-bar pressure, as indicated increase of the Aproj involved in the Piezo1, and recently by the RMSD of the inner pore (Fig. S3 A) and the volume Piezo2, activation (11,14,43). The Aproj can be approxi- of the cavity of the Piezo1 pore region (Fig. S3 B). The value mated by a projected circle on the membrane plane (Fig. 1 of the volume of the cavity in the Piezo1 blade-free con- A). By first approximating the Piezo1 triskelion to the structs is similar to the values observed when Piezo1 was vertices of a scalene triangle, we determined the circum- simulated at 1 bar (Fig. 3 A). We also note that for both con- scribed circle (i.e., the Aproj) as described in Eq. 1 (see Ma- structs (Piezo1D1973 and Piezo1D2104), the APL and the terials and methods). Our stretch simulations are indicative bilayer thickness are comparable with our previous simula- of a concurrent movement of the Piezo1 blades, which all tions at 40 bar (Fig. S2, A and B), but these conditions are flatten synchronously (Videos S1 and S2). Fig. 1 D shows not sufficient to open the pore of a Piezo1 blade-free system. how the increase of applied mechanical force during Moreover, a feature of our blade-free Piezo1 models resides simulations (e.g., more tension) causes an increase of the in their inability to create the Piezo dome (the membrane Aproj value, which reaches 680 nm2 at 40 bar (Fig. 1 indentation), which was suggested to be necessary for the D, orange data). Recent cryo-EM structural data of the Piezo1 mechanosensitivity (11). Overall, our data about full-length Piezo2 estimated a Aproj of the dome to the Piezo1 blade-free constructs show that the blade domain 700 nm2 (43). Although the current value for the full-length 1514 Biophysical Journal 120, 1510–1521, April 20, 2021
Piezo1 opening by membrane tension
FIGURE 2 Coordinated tilting of Piezo1 beams and ion pore helices in response to increased membrane tension. (A) Given is the partial density profile
along the z direction (that is perpendicular to the bilayer x, y plane) for the protein (black line), the membrane (magenta line), and the solvent (i.e., water and
ions; cyan line). Values from the first repeat at 40 bar are shown. Insets are the last frames from simulations, duplicated from Fig. 1 A and for illustrative
purpose only. (B) Shown is a tilt angle relative to the bilayer normal for the beam helix. The transparent regions show the standard deviation. (C) Shown is the
correlation between the RMSD of the Piezo1 blades and the tilt angle relative to the bilayer normal of the TM38 helix. (D) Shown is the correlation between
the distances from the center of mass of the TM38 or TM37 helices and the Piezo1 pore. (E) Shown is the correlation between the RMSD of the Piezo1 blades
and the tilt angle relative to the bilayer normal of the TM37 helix. (F) Given are snapshots of the last frame from the þ1- and 40-bar systems with the Piezo1
shown from the extracellular side. The center is highlighted with a box. The TM38 (cartoon) and the TM37 (ribbon) from each Piezo1 chain are also shown.
Membrane and solvent are not displayed. The color code for the applied negative pressure in each simulation is the same as in Fig. 1. To see this figure in
color, go online.
Piezo1 is still unknown, this agreement suggests that upon inner pore has similar values with the þ1-bar simulation
flattening, our incomplete Piezo1 model covers the area (Fig. S3 B). This suggests that although thinning is observed
that is expected to be within the dome in the closed state during tension application, it is not sufficient to change
calculated for the full-length Piezo2 (43). This is in good Piezo1 shape.
agreement with the suggestion that the dome provides an en- In addition to a structural model of a flat Piezo1, our
ergy store for Piezo1 activation (11). A small increase of stretched simulations allowed us to start investigating how
Aproj was observed in the þ1-bar simulation (Fig. 1 D), sug- the Piezo1 flattened state would adapt once the mechanical
gesting that Piezo1 adapts slightly when inserted in a mem- tension stops. This likely represents a very early postactiva-
brane environment in respect to the cryo-EM structure. tion state. To do that, we extracted the protein coordinates
As discussed above, upon tension application, we from the last frame from our 40-bar systems and carried
observed thinning of the bilayer. To investigate whether out further two all-atom MD simulations of 100 ns each at
bilayer thinning can result in Piezo1 activation, we ran a 1-bar pressure (Piezo140.rep1 and Piezo140.rep2; see Mate-
100-ns simulation with Piezo1 embedded in a DLPC bilayer rials and methods). Calculation of the RMSD of the pore
at 1-bar pressure (Piezo1DLPC; see Table S1 and Materials and of the whole protein (Fig. S5, A and B) shows that for
and methods). In this simulation, the bilayer thickness is both repeats Piezo140.rep1 and Piezo140.rep2, the channel
similar to the values obtained in our 20-bar stretch simu- mouth is still open, and the channel is flat (Fig. S5 C), there-
lation (Fig. S2 B). However, the channel remains closed, fore confirming the stability of our Piezo1 flat structure. In
its blades are curved, and the membrane dome remains as addition, calculation of the Aproj (Fig. 1 D) showed that
it is for the 1-bar system (Fig. S4 C). The RMSD calculated although this value slightly decreased (because no mem-
over the Ca trace for the inner pore and for the full protein brane tension is applied), the Aproj moderately remains
both confirm that the channel remains closed (Fig. S4 A). In above the values obtained for the 30-bar system after
agreement with that, the volume of the cavity of the Piezo1 100 ns.
Biophysical Journal 120, 1510–1521, April 20, 2021 1515
De Vecchis et al.
FIGURE 3 Expansion of the Piezo1 pore region
in response to increased membrane tension. (A)
Shown is the volume of the cavity of the Piezo1
pore region as a function of time for each simulated
system. The applied negative pressure in each simu-
lation is indicated with different colors: black, 1 bar;
red, 5 bar; green, 20 bar; blue, 30 bar; and or-
ange, 40 bar. (B) Shown are the number of water
molecules overlapping with the calculated cavity
in (A) as a function of time for each simulated sys-
tem. (C) Given is the volume of the cavity within
the core of the Piezo1 structure for the þ1- and
40-bar systems. The structures are the last frames
from each simulation and focus only on the Piezo1
pore region. For the 40-bar system, the second
repeat simulation is shown. The cavity (sliced) is
shown as spheres and colored according to the occu-
pancy during the trajectory, from less populated
(blue) to more populated (red). Residues selected
for the calculation of the cavity are shown in ribbon
(including the TM38 helix; see Materials and
methods). TM37, elbow and base helix are repre-
sented as cartoon cylinders. The Piezo1 chains are
shown in the same colors as in Fig. 1 A. (D) Water
molecules (light cyan transparent spheres) within
8 Å from the residues selected for the calculation
of the cavity for the þ1- and 40-bar systems.
The structures are the same shown in (C). (E) Detail
of the Piezo1 channel mouth for the þ1 bar and the
40 bar systems shown from side-on (above) and
extracellular helicopter (below) views. The struc-
tures are the same as presented in (C) and (D) and
show the hydrophobic interactions between the
pore-lining helices TM38 (yellow residues) and the
TM37 (green residues). The salt bridge between
the elbow and the TM38 from a neighboring subunit
is also shown (red and blue residues). Residues from
the Piezo1 hydrophobic gate (residues L2475 and
V2476) are in magenta. To see this figure in color,
go online.
To reconstitute the solvated open pore, because it was As a result, water is no longer able to go through the Piezo1
observed for the 40 bar simulations, lipids that occupied pore. Even though our protocol might have some limita-
the channel mouth were removed after the Piezo140.rep1 tions, e.g., few lipids were removed from the channel
and Piezo140.rep2 system assembly (see Materials and mouth, it suggests that lipids re-enter the channel mouth
methods; Fig. S5 D). Interestingly phospholipids progres- as soon as the tension is removed from the simulations.
sively start to reoccupy the Piezo1 open channel mouth after This may represent a state of Piezo1 in the very early stage
the first 30 ns and finally block the Piezo1 pore (Fig. S5 D). of the inactivation. This might also offer a possible
1516 Biophysical Journal 120, 1510–1521, April 20, 2021
Piezo1 opening by membrane tension
molecular explanation on how lipids re-entering the channel the pore center of mass also shows that larger tension corre-
mouth might contribute to the rapid decay of the Piezo- sponds to a larger displacement (Fig. S6 C). Therefore, dur-
mediated current while the stimulus is still present (47). ing Piezo1 flattening and in-plane expansion, TM38 both
Overall, our approach provides novel molecular insights, increases its tilt angle and its distance from the pore center
to our knowledge, into the activation dynamics of Piezo1 of mass. These two components are correlated with the
channel in a native-like lipid environment. Moreover, we changes in the RMSD calculated for the Piezo1 blades
dynamically quantify the different Aproj expansions under (Fig. 2 C; Fig. S6 C). This shows that TM38 displace-
different mechanical stress and the area covered by a flat, ment/tilting occurs while Piezo1 flattens and expands
activated Piezo1 channel. Our studies also reveal the critical (Fig. 2 F). Similarly, the distance from the pore and tilting
role of Piezo1 blade region in sensing of mechanical force of the TM38 correlates with its antiparallel TM37 (Fig. 2,
and suggest that membrane thinning is not sufficient to acti- D–F; Fig. S6, D and E), suggesting a simultaneous lateral
vate Piezo1 in the absence of tension. Finally, the extended movement and tilting of both helices that might be the
structure obtained at 40 bar agrees with the evidence that mechanism that results in the opening of the pore.
positively correlate the size of the vesicle where Piezo1 is
embedded with a flattened structure (14) and may also be
Tension leads to Piezo1 channel opening
needed for Yoda1 binding (15).
Calculation of the water-filled cavity in Piezo1 pore region
shows that the volume of this cavity increases with the
Structural rearrangements in Piezo1 pore
applied negative pressures in our simulations until it reaches
Piezo1 flattening caused by membrane tension results in the 75,000 Å3 in the 40 bar (Fig. 3 A). The cavity is defined
displacement of the CED relative to the membrane plane by the three TM38 and some regions of the CED (see Mate-
(Fig. 1 A). CED is hidden within the dome at the beginning rials and methods) because the loops that connect this
of the simulations or in the þ1-bar system. Flattening of domain with the TM37/38 form a structure similar to a
Piezo1 and of the bilayer results in the exposure of CED ‘‘cage’’ just above the channel mouth, which is character-
as it emerges outside the dome. The degree of CED expo- ized by a funnel shape (Fig. 3 C). Fig. 3 A shows that tension
sure above the dome increases as the tension in the bilayer of 30 bar, although it contributes to flattening of the
increases; in the 30- and 40-bar systems, CED is fully Piezo1 structure and almost maximizes the beam orienta-
exposed. The density analysis revealed that a considerable tion, it is not sufficient to open the channel pore and maxi-
portion of the Piezo1 structure remained consistently mize the volume of the cavity; the cavity in the 30 bar
exposed to the cytoplasmic side even after the in-plane system is not hydrated as much as in the 40 bar (Fig. 3,
expansion (Fig. 2 A). It is possible that these cytoplasmic re- B and D). Note that in the þ1-bar system, the cavity was
gions remain exposed to the cytosol during activation to pro- not increased, nor was the channel hydrated. After the
vide a platform for Piezo1 interactions with the pore hydration, our 40-bar simulations found that lateral
cytoskeleton, which is in agreement with the cytoskeleton’s fenestrations in the cytosolic region of Piezo1 are also hy-
suggested mechanoprotective role in Piezo1 function (19). drated (Fig. S7 A). Although we did not record complete
Another Piezo1 component that undergoes noticeable ion passage events (likely because of the rather short simu-
structural rearrangement is the beam helix (residues 1300– lation timescale), it is possible that these hydrated fenestra-
1365; Fig. 2 B). During the simulations with tension in the tions may be part of the ion permeation pathway.
bilayer, the tilt angle of each Piezo1 beam changed from Interestingly, the lateral fenestrations were present also at
65 to 85 until it is almost parallel to the bilayer when the beginning of the Piezo140.rep1 and Piezo140.rep2 simu-
Piezo1 is in a flat conformation. The beam tilt movement lations in which we stopped applying tension to the bilayer
is correlated with the change in the RMSD (Fig. S6 A), sug- with the open Piezo1 structure, but they gradually disappear
gesting it may be coupled to Piezo1 flattening. The beam is a because lipids start to reoccupy the Piezo1 channel mouth
long a-helix exposed to the cytoplasm that underpasses the (Fig. S5 D). We note, however, that during our simulations,
three C-terminal proximal bundles. Therefore, our data sug- sodium ions enter the extracellular side of the pore of the
gest that each Piezo1 beam may act as a set of levers and that channel but without completely passing through (see
the concerted movement of the beams during mechanical Fig. S7, B and C).
sensing contributes to gating. We also found that lipid and cholesterol molecules
We next sought to analyze the structural changes initially overlap with this volume, and therefore, they are
involving the pore-lining helices. Calculation of the tilt initially within the Piezo1 pore region when the channel is
angle of the TM38 relative to the bilayer normal shows closed, but they increasingly leave the cavity as higher ten-
that higher negative pressure, and thus larger tension, corre- sion is applied to the bilayer (Fig. 4 A). The three indepen-
sponds to a higher tilting of the pore-lining TM38 (Fig. S6 dent system set-ups followed by as many CG equilibrations
B). The highest tilt angle is observed in the 40-bar system demonstrate that lipids get within the channel mouth during
(Fig. 2 C). Calculation of the distance of TM38 relative to the equilibration phase. This demonstrates that there is
Biophysical Journal 120, 1510–1521, April 20, 2021 1517
De Vecchis et al. FIGURE 4 Proposed mechanism for the Piezo1 channel opening in response to membrane tension. (A) Snapshots from the second repeat simulation at 40 bar with the associated simulation time are shown. Membrane bilayer and solvent (i.e., water and ions) are not shown for clarity. The three Piezo1 subunits are indicated with different colors as in Fig. 1 A. In the extracellular view at the top, the CED has been removed for clarity. Phosphorous and oxygen atoms from lipid headgroups and cholesterol, respectively, are shown (lipids are within 20 Å from the residues selected for the calculation of the cavity; see Materials and methods). The color code is as follows: POPC, tan; POPE, pink; POPS, red; PIP2, gray; and cholesterol, violet. The Ca atoms from interacting residues between the helices TM37 and TM38 and the elbow are indicated as spheres as in Fig. 3 E. In the Piezo1 structures in the middle panel, water molecules within 8 Å from the residues selected for the calculation of the cavity (see Materials and methods) are shown as light cyan transparent spheres. (B) Given is a cartoon representation explaining the Piezo1 activation mechanism coupled to membrane tension. The color code is the same as in (A). The hydrophobic interactions between the TM37 and TM38 and the salt bridge between TM38 and the elbow from a neighboring subunit are indicated in yellow/green and in blue/red, respectively. Lipids that initially occupy the Piezo1 channel mouth and progressively move away from it with the increase tension are sche- matically depicted in pink. Membrane tension and Piezo1 blade expansion are indicated by gray arrows. To see this figure in color, go online. enough space in this closed structure of Piezo1 for lipids to opening. Noticeably, the pore region of the homologous occupy the channel mouth. This may also suggest that the Piezo2 has a similar structure (43), suggesting a similar presence of lipids in the pore may be required to maintain mechanism. In Piezo1, the pore-lining TM38 is located anti- the channel mouth closed (Fig. S5). Indeed, a higher tension parallel to the TM37. The latter, together with the elbow increases the APL in the proximity of the channel mouth region (residues 2116–2142), a base helix (residues 2149– and results in higher volume in the pore cavity because it en- 2175), and hairpin helices (residues 2501–2534) from all ables lipids and cholesterol to move outside of the cavity in three subunits, create a cuff that encloses the pore-lining addition to a larger movement of the pore-lining helices TM38 (Fig. 3 E; (11,14)). In all our simulations, we noticed TM37 and TM38. This hypothesis may explain biochemical two main pivotal interactions: the first is the interaction be- data that suggest a functional role of lipids in Piezo1 regu- tween the TM37 and TM38 via the hydrophobic residues lation (24,48,49). Ile2466, Leu2469, Tyr2470, Ile2473, and Val2477, adjacent Analysis of the lipid distribution around the protein to the hydrophobic functional gate in TM38 involved in (Fig. S8) shows that POPE and POPC lipids in the simula- inactivation (Leu2475 and Val2476 (51)) and the residues tions with stretch applied to the membrane tend to reside Phe2198, Ile2202, Phe2205, and Phe2209 from TM37 within the channel mouth in agreement with our previous (Fig. 3 E). Therefore, extension and expansion of the blades work (50). Therefore, our simulations suggest that the ten- cause the displacement of TM37 that in turn ‘‘pulls’’ the sion applied to the bilayer during functional activation re- pore-lining helix TM38 because of hydrophobic interac- sults not only in the displacement of TM38 and TM37 but tions. We suggest that this results in the opening of the chan- also in the movement of lipids away from the Piezo1 pore. nel and might explain why in the 40-bar simulation in These two events occur simultaneously and may both be which more expansion was observed, the pore was more required for Piezo1 full opening. open (Fig. 3 E). An open question is how the mechanical stimuli are trans- The second pivotal interactions occur in a region toward ferred from the blades to the channel pore. Our results sug- the C-terminal end of TM37 and TM38. Previous work sug- gest that the flattening and expansion of the blades is gested the importance of the negative charge at position transmitted to the TM helices 37 and 38, guiding the pore Glu2133 (Glu2416 in Piezo2) for conductance and ion 1518 Biophysical Journal 120, 1510–1521, April 20, 2021
Piezo1 opening by membrane tension
selectivity (52). In particular, the single mutant E2133A sively move away from the channel pore, opening the
showed half the unitary conductance with respect to the Piezo1 channel. And 4) in the open conformation, Piezo1
wild-type phenotype, and the negative charge from the blades are fully extended and flat, CED is exposed, and
glutamate has been suggested to allosterically modulate beam helices are almost parallel to the lipid bilayer. The
the selectivity of the Piezo1 filter (52). The Piezo1 cryo- elbow regions from neighboring subunits pull the pore-lin-
EM structure maps the position Glu2133 in a helix named ing helices, TM38. The increase in the APL due to tension
the elbow. Interestingly, the tip of each elbow points toward allows displacement of the lipids that occupy the channel
the TM38 of another Piezo1 subunit, and in our model, we mouth, resulting in an open conformation. Therefore, both
found that Glu2133 on the elbows forms a salt bridge with conformational changes and increase in the APL that allow
the Arg2482 in a neighboring subunit in agreement with pre- movement of lipids away from the channel pore region
vious suggestions (Fig. 3 E; (12)). This salt bridge is re- might be required for Piezo1 channel ion pore opening.
tained in all our simulations between at least two Piezo1 Our studies also demonstrate that it is the blade region
chains with above 70% of occurrence, even after Piezo1 that senses the tension in the membrane and that membrane
conformational rearrangement because of the applied nega- thinning is not sufficient to activate Piezo1 without the pres-
tive pressure (Table S2). Therefore, we suggest that this ence of tension.
interface is maintained during the conformational transition. Recent structural data showed that the second member of
In agreement with this hypothesis, functional data (52) this family of proteins, Piezo2, has a similar shape with
demonstrated that the mutant E2133D increases the conduc- Piezo1. However, Piezo1 and Piezo2 have rather low
tance with respect to the wild-type, which is probably due to sequence similarity, and Piezo2 is larger. Additionally,
the short chain of the aspartate that would pull TM38 back- Piezo2 is likely to create a deeper dome in the membrane
ward toward the elbow, thus reducing the pore constriction. (9 nm compared with 6 nm of Piezo1) and have a larger
Furthermore, this hypothesis rationalizes the effect of the footprint because of its larger size. The two channels also
less efficient mutant E2133K that was shown to decrease have differences in the topology of their pore region. There-
Piezo1 conductance (52). That is, the mutation to lysine fore, although it is possible that the mechanism presented
opposite to the Arg2482 would push forward the TM38 here may be similar for Piezo2, further studies are needed
because of repulsion, contributing to the increase of the to examine whether differences in Piezo1 and Piezo2 will
channel constriction. In the E2133K mutant, the salt bridge result in differences in activation and sensitivity to tension.
is broken and so is the connection between the elbow and A recent study has also suggested that other activation
the TM38, which results in an impairment of the functional mechanisms may be possible for Piezo1 (53,54). In partic-
displacement connected to the blade expansion. In this con- ular, Piezo1 was suggested to be an RNA sensor for sin-
dition, although less conducting, the channel is still func- gle-strand RNA (ssRNA) (53). It is still unclear whether
tional (52), which may be due to the aforementioned RNA binds to Piezo1, and thus, it is not possible to know
hydrophobic interactions between TM37 and TM38 in syn- how ssRNA might activate the channel. However, our recent
ergy with the proposed lipid displacement. In conclusion, work predicts strong interactions between Piezo1 and
the predicted salt bridge between Glu2133 with Arg2482 anionic phospholipids present exclusively in the inner leaflet
in a neighboring subunit would provide an interaction inter- of the membrane (50). It may be possible that the ssRNA
face between the Piezo1 chains so that each blade extension backbone negative charge might have a potential similar
would control the gating of a nearby Piezo1 chain (Fig. 4 B). role to anionic lipids on Piezo1 dynamics, but further studies
This may also provide a mechanism that ensures a prompt are needed to understand how ssRNA activates Piezo1.
and mutual propagation of sensing between Piezo1 subunits. It has also been shown that the small molecule Yoda1 can
activate Piezo1. A recent study proposed a molecular-wedge
mechanism for Yoda1-mediated Piezo1 activation (15).
CONCLUSIONS
Therefore, it is possible that functional Yoda1 binding sites
This study reveals possible molecular principles by which might become available during the conformational changes
physiological Piezo1 channels may open in response to associated to the flattening observed in our study. Moreover,
increased membrane tension. The suggested steps are shown Yoda1 is a hydrophobic molecule, and thus, it is likely to
in Fig. 4 B and are as follows: 1) without applied tension, interact with the membrane bilayer, possibly altering the
Piezo1’s curved shape forces the bilayer to bend inwards, thickness and/or the membrane curvature. These alterations
creating a stable indented trilobed topology with the CED could have a local effect in the proximity of the protein
hidden within it. In this resting conformation, the Piezo1 affecting Piezo1 flattening. Further studies are needed to
channel mouth is occupied by lipids. 2) Membrane tension clarify the molecular mechanism of how Yoda1 may acti-
on the bilayer results in the flattening and in-plane expan- vate Piezo1.
sion of its blades and tilting of its beam helices. 3) Piezo1 Although this model informs ideas about the Piezo1 chan-
expansion and flattening allows the blades to ‘‘pull’’ helices nel’s response to force and its consequent opening and it
37 and 38 via hydrophobic interactions to tilt and progres- agrees with previous experimental evidence and hypotheses,
Biophysical Journal 120, 1510–1521, April 20, 2021 1519De Vecchis et al.
it nevertheless has potential limitations that should be 4. Fotiou, E., S. Martin-Almedina, ., P. Ostergaard. 2015. Novel muta-
tions in PIEZO1 cause an autosomal recessive generalized lymphatic
considered. The modeled channel is a fragment devoid of dysplasia with non-immune hydrops fetalis. Nat. Commun. 6:8085.
three N-terminal bundles in each Piezo1 and long, probably 5. Andolfo, I., S. L. Alper, ., A. Iolascon. 2013. Multiple clinical forms
unstructured, cytoplasmic loop regions. Whereas the latter of dehydrated hereditary stomatocytosis arise from mutations in
are still poorly annotated, Piezo1 blades have been associ- PIEZO1. Blood. 121:3925–3935.
ated with pressure sensing, thus at this stage, we are not 6. Ma, S., S. Cahalan, ., A. Patapoutian. 2018. Common PIEZO1 allele
in african populations causes RBC dehydration and attenuates plasmo-
able to fully extend the model to the complete protein. dium infection. Cell. 173:443–455.e12.
Additionally, as mentioned above, usage of nonphysio- 7. Li, J., B. Hou, and D. J. Beech. 2015. Endothelial Piezo1: life depends
logical pressure or temperature may induce some caveats. on it. Channels (Austin). 9:1–2.
Moreover, Piezo1 is activated by shear stress along the 8. Coste, B., J. Mathur, ., A. Patapoutian. 2010. Piezo1 and Piezo2 are
membrane surface, an important component not present in essential components of distinct mechanically activated cation chan-
nels. Science. 330:55–60.
our model, which could be additive to the stretch in vivo.
9. Ge, J., W. Li, ., M. Yang. 2015. Architecture of the mammalian me-
Further, Piezo1 functions at the millisecond timescale, chanosensitive Piezo1 channel. Nature. 527:64–69.
which in conjunction with the recorded activating pressure, 10. Zhao, Q., H. Zhou, ., B. Xiao. 2018. Author correction: structure and
led us to a further limitation when using computer simula- mechanogating mechanism of the Piezo1 channel. Nature. 563:E19.
tions. However, we were able to accelerate the process 11. Guo, Y. R., and R. MacKinnon. 2017. Structure-based membrane dome
and investigate for the first time, to our knowledge, not mechanism for Piezo mechanosensitivity. eLife. 6:e33660.
only a flatten Piezo1 but intermediate conformations from 12. Saotome, K., S. E. Murthy, ., A. B. Ward. 2018. Structure of the me-
chanically activated ion channel Piezo1. Nature. 554:481–486.
the closed to open state within a simulated native lipid
environment. 13. Haselwandter, C. A., and R. MacKinnon. 2018. Piezo’s membrane
footprint and its contribution to mechanosensitivity. eLife. 7:e41968.
Despite the considerations above regarding pressure dy-
14. Lin, Y.-C., Y. R. Guo, ., S. Scheuring. 2019. Force-induced conforma-
namics, we believe that this study informs how the fasci- tional changes in PIEZO1. Nature. 573:230–234.
nating fold and shape of Piezo1 channel might allow this 15. Botello-Smith, W. M., W. Jiang, ., Y. Luo. 2019. A mechanism for the
protein to sense and transduce mechanical signals activation of the mechanosensitive Piezo1 channel by the small mole-
throughout the membrane, and we hope it will help to cule Yoda1. Nat. Commun. 10:4503.
pave the way for future investigation once the Piezo1 full- 16. Chiang, C.-S., A. Anishkin, and S. Sukharev. 2004. Gating of the large
mechanosensitive channel in situ: estimation of the spatial scale of the
length structure becomes available. transition from channel population responses. Biophys. J. 86:2846–
2861.
17. Teng, J., S. Loukin, ., C. Kung. 2015. The force-from-lipid (FFL)
SUPPORTING MATERIAL principle of mechanosensitivity, at large and in elements. Pflugers
Arch. 467:27–37.
Supporting material can be found online at https://doi.org/10.1016/j.bpj. 18. Martinac, B., J. Adler, and C. Kung. 1990. Mechanosensitive ion chan-
2021.02.006. nels of E. coli activated by amphipaths. Nature. 348:261–263.
19. Cox, C. D., C. Bae, ., B. Martinac. 2016. Removal of the mechano-
protective influence of the cytoskeleton reveals PIEZO1 is gated by
AUTHOR CONTRIBUTIONS bilayer tension. Nat. Commun. 7:10366.
20. Poole, K., M. Moroni, and G. R. Lewin. 2015. Sensory mechanotrans-
The research was designed by D.D.V., D.J.B., and A.C.K. D.D.V. per- duction at membrane-matrix interfaces. Pflugers Arch. 467:121–132.
formed the research and acquired and analyzed all the data. A.C.K. and 21. Jefferies, D., J. Shearer, and S. Khalid. 2019. Role of O-antigen in
D.J.B. supervised the project. D.D.V., D.J.B., and A.C.K. wrote the article. response to mechanical stress of the E. coli outer membrane: insights
from coarse-grained MD simulations. J. Phys. Chem. B. 123:3567–
3575.
ACKNOWLEDGMENTS 22. Kanda, H., and J. G. Gu. 2017. Membrane mechanics of primary
afferent neurons in the dorsal root ganglia of rats. Biophys. J.
A.C.K. and D.D.V. are funded by an Academy of Medical Sciences and 112:1654–1662.
Wellcome Trust Springboard Award (SBF002\1031). D.J.B. is funded by 23. Ridone, P., M. Vassalli, and B. Martinac. 2019. Piezo1 mechanosensi-
a Wellcome Trust Investigator Award (110044/Z/15/Z). tive channels: what are they and why are they important. Biophys. Rev.
11:795–805.
24. Pliotas, C., and J. H. Naismith. 2017. Spectator no more, the role of the
membrane in regulating ion channel function. Curr. Opin. Struct. Biol.
REFERENCES 45:59–66.
25. Sali, A., and T. L. Blundell. 1993. Comparative protein modelling by
1. Wu, J., A. H. Lewis, and J. Grandl. 2017. Touch, tension, and transduc- satisfaction of spatial restraints. J. Mol. Biol. 234:779–815.
tion - the function and regulation of Piezo ion channels. Trends Bio-
chem. Sci. 42:57–71. 26. Fiser, A., R. K. Do, and A. Sali. 2000. Modeling of loops in protein
structures. Protein Sci. 9:1753–1773.
2. Beech, D. J., and A. C. Kalli. 2019. Force sensing by Piezo channels in 27. Shen, M. Y., and A. Sali. 2006. Statistical potential for assessment and
cardiovascular health and disease. Arterioscler. Thromb. Vasc. Biol. prediction of protein structures. Protein Sci. 15:2507–2524.
39:2228–2239.
28. Abraham, M. J., T. Murtola, ., E. Lindahl. 2015. GROMACS: high
3. Li, J., B. Hou, ., D. J. Beech. 2014. Piezo1 integration of vascular ar- performance molecular simulations through multi-level parallelism
chitecture with physiological force. Nature. 515:279–282. from laptops to supercomputers. SoftwareX. 1–2:19–25.
1520 Biophysical Journal 120, 1510–1521, April 20, 2021You can also read
